README.md

metadata

base_model: seyonec/ChemBERTa-zinc-base-v1
library_name: transformers
license: mit
tags:
  - PROTAC
  - cheminformatics
  - generated_from_trainer
model-index:
  - name: ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25
    results: []

ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25

This model is a fine-tuned version of seyonec/ChemBERTa-zinc-base-v1 on the ailab-bio/PROTAC-Splitter-Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.3124
• E3 Graph Edit Distance Norm: inf
• Poi Has Attachment Point(s): 0.9294
• Linker Heavy Atoms Difference: 0.3252
• Reassembly Nostereo: 0.5845
• E3 Valid: 0.9942
• Poi Heavy Atoms Difference Norm: 0.0668
• Linker Graph Edit Distance: inf
• All Ligands Equal: 0.5477
• Has All Attachment Points: 0.9857
• Linker Valid: 0.9951
• E3 Tanimoto Similarity: 0.0
• Heavy Atoms Difference: 6.4929
• Tanimoto Similarity: 0.0
• Reassembly: 0.5549
• E3 Heavy Atoms Difference: 0.3628
• Poi Graph Edit Distance Norm: inf
• Valid: 0.9232
• Linker Tanimoto Similarity: 0.0
• Linker Heavy Atoms Difference Norm: 0.0050
• Poi Valid: 0.9294
• Linker Graph Edit Distance Norm: inf
• Poi Equal: 0.7673
• Linker Equal: 0.7726
• E3 Graph Edit Distance: inf
• Poi Graph Edit Distance: inf
• Has Three Substructures: 0.9983
• Poi Heavy Atoms Difference: 2.0849
• Num Fragments: 3.0008
• Poi Tanimoto Similarity: 0.0
• E3 Heavy Atoms Difference Norm: 0.0044
• E3 Has Attachment Point(s): 0.9942
• E3 Equal: 0.8035
• Linker Has Attachment Point(s): 0.9951
• Heavy Atoms Difference Norm: 0.0854

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 128
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 699
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	E3 Graph Edit Distance Norm	Poi Has Attachment Point(s)	Linker Heavy Atoms Difference	Reassembly Nostereo	E3 Valid	Poi Heavy Atoms Difference Norm	Linker Graph Edit Distance	All Ligands Equal	Has All Attachment Points	Linker Valid	E3 Tanimoto Similarity	Heavy Atoms Difference	Tanimoto Similarity	Reassembly	E3 Heavy Atoms Difference	Poi Graph Edit Distance Norm	Valid	Linker Tanimoto Similarity	Linker Heavy Atoms Difference Norm	Poi Valid	Linker Graph Edit Distance Norm	Poi Equal	Linker Equal	E3 Graph Edit Distance	Poi Graph Edit Distance	Has Three Substructures	Poi Heavy Atoms Difference	Num Fragments	Poi Tanimoto Similarity	E3 Heavy Atoms Difference Norm	E3 Has Attachment Point(s)	E3 Equal	Linker Has Attachment Point(s)	Heavy Atoms Difference Norm
0.0086	0.4932	5000	0.2931	inf	0.9248	0.6527	0.5238	0.9949	0.0785	46033994334277620957572376380209976047687534236138629465374720.0000	0.4899	0.9820	0.9954	0.0	7.4350	0.0	0.4972	0.3592	inf	0.9182	0.0	0.0200	0.9248	0.0544	0.7416	0.7092	inf	inf	0.9988	2.4199	3.0004	0.0	0.0041	0.9949	0.7832	0.9954	0.0974
0.004	0.7398	7500	0.3098	inf	0.9225	0.2498	0.5729	0.9924	0.0709	59313031161473097104973641187183678565005524270457129338404864.0000	0.5381	0.9818	0.9941	0.0	6.9868	0.0	0.5451	0.4005	inf	0.9147	0.0	-0.0010	0.9225	inf	0.7603	0.7604	inf	inf	0.9978	2.2024	3.0006	0.0	0.0081	0.9924	0.8044	0.9941	0.0924
0.003	0.9864	10000	0.3124	inf	0.9294	0.3252	0.5845	0.9942	0.0668	inf	0.5477	0.9857	0.9951	0.0	6.4929	0.0	0.5549	0.3628	inf	0.9232	0.0	0.0050	0.9294	inf	0.7673	0.7726	inf	inf	0.9983	2.0849	3.0008	0.0	0.0044	0.9942	0.8035	0.9951	0.0854

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1